[論文レビュー] Impact of aberrations in SLM-based far-field holography

Markus Zimmermann, Andreas Brenner+2 more|arXiv (Cornell University)|Feb 4, 2026

Advanced Optical Imaging Technologies被引用数 0

ひとこと要約

The paper uses camera-in-the-loop calibration with a neural-network–driven digital twin to assess how different modeled aberrations affect image quality, diffraction efficiency, and speckle in far-field holography, showing phase aberrations have the strongest impact and that including all aberrations yields the best metrics.

ABSTRACT

We use camera-in-the-loop calibration to calibrate a phase-only spatial light modulator (SLM) in a far-field hologram setup. The recorded intensity distributions achieve a high degree of consistency with the calculated results, indicating a precise calibration and sufficient modeling of the most prominent aberrations. In this work, we discuss the modeled aberrations and examine the improvement or loss in image quality and diffraction efficiency that is obtained by including or excluding the modeled aberrations in the calibration. We further show the influence of aberrations on speckle-reduced holograms and evaluate the speckle contrast.

研究の動機と目的

• Evaluate how spatially varying illumination, fringing field effects, non-linear phase modulation, phase aberrations, and illumination amplitude influence holographic reconstruction.
• Quantify improvements in image quality and speckle reduction when aberrations are included in a CITL calibration.
• Identify which aberrations most strongly affect PSNR, SSIM, speckle contrast, and diffraction efficiency.

提案手法

• Use camera-in-the-loop (CITL) calibration with a physically interpretable neural network (digital twin) where each layer corresponds to a specific optical aberration.
• Incorporate modeled aberrations: phase modulation, fringing field effect, phase aberrations (low and high spatial frequencies), illumination amplitude, and fill-factor/pixel shape.
• Calibrate the system by minimizing mean squared error between predicted and recorded far-field intensity.
• Compare calibration models with different aberrations excluded to assess impact on image quality and diffraction efficiency.
• Evaluate speckle contrast using a specified area and a speckle suppression method.
• Measure diffraction efficiency from a blazed grating and the zero-order when nothing is displayed.

実験結果

リサーチクエスチョン

• RQ1 How do boundary conditions and modeled aberrations influence the accuracy of far-field hologram reconstructions?
• RQ2 What is the relative importance of phase aberrations versus other aberrations on image quality and speckle?
• RQ3 Can CITL calibration compensate for certain aberrations, and how does excluding them affect PSNR, SSIM, speckle, and efficiency?
• RQ4 How do modeled aberrations alter the learned phase modulation, fringing-field kernels, and illumination amplitude?
• RQ5 What is the effect of different aberration sets on diffraction efficiency and speckle reduction?

主な発見

plain SGD	w/o fringing field effect	linear phase mod.	w/o phase aberrations	uniform illumination	full model
PSNR	14.90	20.24	20.67	15.69	19.76	21.23
SSIM	0.16	0.22	0.24	0.16	0.27	0.26
C	0.45	0.15	0.14	0.37	0.15	0.11
η	0.68	0.72	0.70	0.71	0.70	0.70

• Including all aberrations yields the best PSNR, SSIM, speckle reduction, and diffraction efficiency.
• Phase aberrations have the strongest influence on image quality, with defocus and primary spherical aberration being most prominent among assessed terms.
• Fringing field effect learns non-uniform 3×3 kernels that reflect pixel crosstalk, with diagonal elements near zero and center-dominant values.
• Illumination amplitude has the lowest effect on image quality, with uniform illumination sometimes yielding higher SSIM.
• Diffraction efficiency is not maximized when the fringing field is modeled to maximize efficiency in the optimization setup, suggesting model-calibration dynamics differ from direct optimization.
• The full model (all aberrations) achieves the highest PSNR and lowest speckle contrast.